AI-ACCELERATED DRUG DISCOVERY
Selective Inhibitors for one of six Ion Channel Isoforms
Mitigating off-target effects within higly similar ion channels family

Background

  • The family of highly similar ion channel isoforms.
  • Only one of which have to be targeted.
  • The exact location of the most favorable binding pocket for achieving selectivity is unknown.
  • The goal is to design highly selective inhibitors against 6 ion channel isoforms.

Methodology

  • Three tentative binding sites were identified for each protein isoform by proprietary Receptor.AI pocket detection workflow:
    • in the outer channel pore,
    • inside the channel cavity,
    • between the functionally important transmembrane helices.
  • Selectivity assessment based on differential pocket pharmacophore’s representation combined with generative AI binding pose prediction was used.
  • Pre-filtered stock chemical space of 662K compounds was used as well as custom focused diversity database of 50K compounds.
  • 291 compounds were selected for experimental validation.

Results

  • Three metrics were used:
    • Fold increase of effect on target isoform in comparison to off-target isoforms.
    • UFD effect: the preference of the compound to block the active channel state relative to the resting state.
    • Peak blocking of the target channel at 120 μM of compound relative to vehicle.
  • The best compound have shown x3.16 selectivity against the target isoform.
  • Top 5 compounds demonstrate >x2 selectivity and specificity to active state of the target channel, which demonstrates correct mechanism of action against the active channel conformation.
  • Top 10 compounds are attributed to all three binding pockets.
  • The top 3 compounds was tested for activity in vivo (murine model) and are proven to be active (including a highly potent one) and non-toxic.